Determining the green vegetation fraction from RapidEye data for use in regional climate simulations

Transcription

1 Research Unit 1695 Determining the green vegetation fraction from RapidEye data for use in regional climate simulations Kristina Imukova, Joachim Ingwersen and Thilo Streck Institute of Soil Science and Land Evaluation University of Hohenheim

2 The Research Unit ( Forschergruppe ) 1695 The present study is a part of the Research Unit 1695 Structure and Functions of Agricultural Landscapes under Global Climate Change - Processes and Projections on a Regional Scale, in short Regional Climate Change. Overall goals of the Research Unit: q to assess the impact of climate change on structure and functions of agricultural landscapes on a regional scale q to improve existing models and to combine them into an integrated land system model

3 Green Vegetation Fraction () in NOAH-MP Land Surface Model (LSM) q NOAH-MP LSM is a part of the weather and climate models. It simulates land surface exchange processes. q is a regular space-time gridded input to evapotranspiration scheme in the NOAH-MP LSM. is used for partitioning evapotranspiration into direct soil evaporation and canopy transpiration. Source: Niu et al., 211

4 Remote sensing of the Green Vegetation Fraction () Objectives: q to develop a valid calibration scheme to derive the from RapidEye satellite images for Kraichgau region based on ground truth measurements q to generate high-resolution gridded data for this region in a monthly resolution q to quantify the temporal and spatial variability of the of croplands at the field and the regional scale q to improve the description of the vegetation dynamics of croplands in the NOAH-MP LSM

5 The ground truth measurement of the Three fields in Kraichgau Five subplots 1 1 m 2 per field Study crops: winter wheat, winter rape and silage maize Nikon Coolpix P7 with automatic settings Resolution of photo: pixels ^ _!! Sinsheim Pforzheim! Stuttgart ± km km

6 The ground truth measurement of the Three fields in Kraichgau Five subplots 1 1 m 2 per field Study crops: winter wheat, winter rape and silage maize Nikon Coolpix P7 with automatic settings Resolution of photo: pixels

7 Image processing Rundquist et al. (22) Pixel value < -1 -> green - = Red band (-255) Green band (-255) Contrast raster

8 Date (DOY) 1/4 (11) 2/5 (123) 1/6 (153) 8/6 (16) 22/6 (174) 29/6 (181) 6/7 (188) 13/7 (195) 2/7 (22) 1/8 (223) 17/8 (23) 22/8 (235) 5/9 (249) 14/9 (258) Maize Wheat Rape Maize Wheat Rape Date (DOY) Green vegetation fraction in percentage

9 Dynamics of of the study crops 1. Winter Rape Maize Winter Wheat - ground truth Date, year Date, year 213

10 Comparison of ground truth data with satellite data for Baden-Württemberg region 1. from NESDIS/NOAA satellite data (~15x15 km 2 ) from ground truth measurements (~1x1 m 2 ) (agricultural area) May June July August Month

11 Remote sensing of the green vegetation fraction Estimation of the from NDVI data NDVI map derived from RapideEye satellite image ( ) Calculation of according to Gutman and Ignatov (1998) 1. NDVI = (NIR-R)/( NIR+R) 2. = (NDVI-NDVI )/(NDVI - NDVI ) For our study sites this equation reads: = 1.11NDVI -.6

12 Relation between and NDVI satellite data 1..8 Gutman & Ignatov approach GVI=1.11NDVI-.6 EF=.86 - Groundtruth GVI=1.15NDVI-.16 EF=.91 EF model efficiency NDVI RapidEye

13 Validation against the second independent data set 1. - Calibration EF model efficiency EF =.91 RMSE =.11 bias = Groundtruth

14 Map of Kraichgau region 4, Histogram map May 3, 2, 1, value is used in NOAH LSM, is derived from NESDIS/NOAA satellite data ~15x15 km 2

15 Histogram of the of croplands 3 24 May, R 2 = June, 212 R 2 = July, 212 R 2 = September, 212 R 2 = October, 212 R 2 = April, 213 R 2 = May, 213 R 2 = July, 213 R 2 = August, 213 R 2 = September, 213 R 2 =

16 dynamics of croplands at the regional scale 1. Early covering crops Late covering crops Apr May Jun Jul Aug Sep Month

17 Conclusions and Outlook: q The Green Vegetation fraction (), required by land surface models, can be estimated well at high resolution from NDVI retrieved from RapidEye images q The distribution of the of croplands is highly variable both in space and time q At the regional scale, of croplands show a distinct bi-modal distribution due to the pronounced phenological differences between early covering (winter wheat, winter rape etc.) and late covering crops (maize, sugar beet) q It is to be expected that subdividing the crop functional type in LSM into early and late covering crops will considerably improve the simulation of water and energy fluxes at the land surface q In future, we will study the effect of five aggregation levels of maps (5x5 m 2, 1x1 m 2, 1x1 km 2, 5x5 km 2, and 15x15 km 2 ) on the simulation of surface energy fluxes with the NOAH-MP land surface model

18 Thank you for your attention